Hi

I am running a channel flow using LES. My problem is that turbulence after developing for a bit starts to decay in channel. And after a while I see zero turbulence in my domain. What can I do under such circumstances to sustain the turbulence. Thank you for your answer.

You will have to give details. What is the Re? What is the grid size? What is the order of accuracy for spatial/temporal derivatives? What is the initial condition?

Hi agg

Thank you for your answer. I am using CFX for this pipe flow (well I am running both pipe and channel flow and same thing is happening in both cases). I am using periodic boundary and have set pressure drop between inlet and outlet.

My reynolds number at the start of simulation was 40000, however my velocity keeps on rising and I suspect by now its much higher than 50000. Length of my pipe is 1.6m and diameter is 0.1m. I have half a million cells in my grid, and I have put lots of cells in radial direction to resolve the bondary layer. CFX has 2nd order accuracy and uses central difference scheme for discrestisation. I am using RANS model as initial guess.

Do I have to specify pressure gradient as well when I use periodic boundary? I will be thankful for any hints/suggestions which would help to sustain turbulence in the pipe so that I can get fully developed flow. Thank you

It doesn't make sense that your turbulence decays and at the same time velocities rise. This is counter intuitive.

Maintaining streamwise pressure gradient is extremely important even with periodic bc in streamwise direction. This is done via source term in the NS equation. Look up papers for more information

I think in your simulations, the initial pressure gradient is not being maintained, resulting in pressure drop and therefore loss of turbulence

Hi agg

I think I have sorted the problem. I was observing zero turbulence near walls which made me think that turbulence is decaying. When I increased the mesh resolution near walls and resolved the boundary layer, the problem disappeared. Thank you for your help.

[mixer]

Hi CFDls

Have you reported your CFX LES with results Pipe flow results anywhere ( a publication or a technical report). I would like to have a look at it.

[aatman]

Originally Posted by agg
;55803

It doesn't make sense that your turbulence decays and at the same time velocities rise. This is counter intuitive.

No, this have sense! if turbulence decay there are no turbulent transfer of momentum in the wall normal as in the spanwise direction and with a fix pressure gradient velocity must increase in the streamwise direction.

I think that if he is making "LES" with RANS model for turbulence this has no sense. He is providing a very high eddy viscosity bringing the flow to laminarization.